Novel swirl flow-focusing microfluidic device for the production of monodisperse microbubbles

(will be inserted by the editor)

Novel swirl flow-focusing microfluidic device for the production

of monodisperse microbubbles

Irene Arcos-Turmo · Miguel ´Angel Herrada · Jos´e Mar´ıa L´opez-Herrera · David Fernandez Rivas · Alfonso M. Ga˜n´an-Calvo · Elena

Castro-Hern´andez

Received: date / Accepted: date

Abstract A novel swirl flow-focusing microfluidic axisym-metric device for the generation of monodisperse micro-bubbles at high production rates to be used as in-line contrast agents for medical applications is presented. The swirl effect is induced upstream of the discharge orifice by a circular array of microblades which form a given angle with the radial direction. The induced vorti-cal component on the focusing liquid stabilizes the gas meniscus by the vorticity amplification due to vortex stretching as the liquid is forced through the discharge orifice. The stabilized meniscus tapers into a steady gas ligament that breaks into monodisperse microbubbles. A reduction up to 57% in the microbubble diameter is accomplished when compared to conventional axisym-metric flow-focusing microdevices. An exhaustive ex-perimental study is performed for various blade angles and numerous gas to liquid flow rate ratios, validating previous VoF numerical simulations. The microbubbles issued from the stabilized menisci verify prior scaling law of flow-focusing.

Keywords microbubble · flow-focusing · swirl

Irene Arcos-Turmo · Jos´e Mar´ıa L´opez-Herrera · Alfonso M. Ga˜n´an-Calvo · Miguel ´Angel Herrada · Elena Castro-Hern´andez

´

Area de Mec´anica de Fluidos, Departamento de Ingenier´ıa Aeroespacial y Mec´anica de Fluidos, Universidad de Sevilla, Avenida de los Descubrimientos s/n 41092, Sevilla, Spain. E-mail: elenacastro@us.es

David Fern´andez Rivas

Mesoscale Chemical Systems and MESA+_{Institute of}

Nano-technology, P.O. Box 217, 7500 AE Enschede, The Nether-lands.

Fig. 1 SEM image of the 60◦ blade swirl flow-focusing mi-crofluidic device.

1 Introduction

Microbubbles represent not only the obvious counter-part of sprays to diffuse a fluid phase enclosed by sur-face tension into the environment. Here, the environ-ment is a liquid. Given its enormous inertia combi-ned with high surface tension forces at small scales, the generation, manipulation and dynamics of micro-bubbles possess unique features not found in any other system. The behavior of microbubbles is often counte-rintuitive, and generally nonlinear (Marmottant et al, 2005). Thus, the physics involved is drastically domina-ted by very large inertia effects of the environment and the compressibility of the microbubbles. That compres-sibility combined with their surface properties make them perfect devices for some critical tasks: in bio-medical applications, as contrast agents, or as vehicles for drug delivery, or gene therapy by sonoporation (Ta-kahashi, 2005; Ferrara et al, 2007).

As a proxy to qualify the importance of microbub-bles as established devices or tools in medicine, one may compare the relative percentages of scientific

arti-cles published mentioning certain combinations of key-words. For example, one may use public databases to assess that the percentage of papers mentioning “ul-trasound”, and “microbubbl*” or “micro-bubbl*” over the total has stabilized around 0.024% over the last ten years. The vast majority of them use the ultrasound-microbubble coupled dynamics as the key device to produce benefits in “cardio*” or “cancer*” related ap-plications (Wei et al, 1998; Kennedy et al, 2004). For example, this is comparable to the weight of “amioda-rone” , the most widely used antiarrhythmic drug, in cardiology publications (0.05% of papers over the last five years), which illustrates the social and economic importance of microbubble-contrast agents in certain medical fields. In this regard, one may verify that the number of papers mentioning “cardio*” and “cancer*” have stabilized at 3.5% and 6.8%, respectively, over the last hundred years, making them the highest concerns of medicine since long ago. For applications in these fields as contrast agents, to achieve the highest possi-ble efficiency and to minimize gas infusion and adverse effects, reducing the microbubble size and its disper-sion as much as possible is of paramount importance: a focused, single frequency excitation is the best way to manipulate swarms of microbubbles in a liquid (gene-rally, blood) stream.

Thus, the search of physical principles and develop-ment of technologies to produce the highest possible surface (or minimum microbubble size) per unit input energy, concentrated around a single size value (mo-nodispersity), has fueled an immense collective effort. Almost unfailingly, the solutions proposed for the effi-cient one-step generation of microbubbles make use of microfluidic designs. In general terms, microfluidics has co-evolved driven by strong expectations in the areas of biomedicine and new materials (Whitesides, 2006). Here, microfluidic devices have demonstrated to be an attractive method to mass-produce narrowly distribu-ted micron size microbubbles in a wide range of liquids (Ga˜n´an-Calvo and Gordillo, 2001; Anna et al, 2003; Garstecki et al, 2006). Several microfluidic techniques have been developed in the last decade, being the T jun-ctions, cross junctions and flow-focusing designs those with the strongest boost.

Despite their differences, two global regimes can be identified: (i) a bubbling regime, where bubbles are for-med right at the tip of the injection tube (axisym-metric case) or at the entrance of the outlet channel (planar case), and (ii) a jetting regime characterized by the generation of a jet that breaks up into bubbles (Ga˜n´an-Calvo and Gordillo, 2001; Ga˜n´an-Calvo, 2004; Ga˜n´an-Calvo et al, 2006). These designs can be imple-mented in axisymmetric geometries (Ga˜n´an-Calvo and

Gordillo, 2001), by means of concentric capillary tu-bes, or in planar configurations using techniques such as soft lithography or micromachining. Bubbles gene-rated by planar flow-focusing devices (Garstecki et al, 2004, 2005) and T junctions (Garstecki et al, 2006; Do-llet et al, 2008) have usually diameters of the order of the device geometry as a result of the pinch-off process of either the bubbling or squeezing regimes. It is by for-cing both, the liquid and gas streams, through a small aperture or constriction, using flow-focusing (Ga˜n´ an-Calvo and Gordillo, 2001) or cross junctions (Castro-Hern´andez et al, 2011), that it is possible to achieve bubbles whose size is considerable smaller than the cha-racteristic geometric length. Flow-focusing ensures the production of monodisperse microbubbles at high and controlled production scales. The strong focusing effect created at the constriction induces the formation of a steady tapering gas meniscus, from which bubbles are ejected.

In this work, we propose a robust one-step met-hod to controllably produce small monodisperse mi-crobubbles in an aqueous liquid stream at high pro-duction rate, to be employed -among other uses- as contrast agents for medical applications: a novel swirl flow-focusing (SFF) microfluidic device. The essential geometrical difference of our device with respect to all previous implementations is the presence of a circular blade array, concentric to the exit channel, that for-ces the liquid to swirl around the exit hole (see Figure 1). The centrifugal forces created by the swirl origi-nate an intense pressure gradient in the radial direc-tion, stabilizing the gas meniscus and focusing the gas into a short steady gas ligament. Thus, the imposed swirl enables the formation of a tapering meniscus for a wider range of working experimental conditions than in common flow-focusing. This extends the robustness and versatility of co-flow designs to mass produce very small microbubbles to inaccessible parametrical ranges to other known configurations.

2 Materials and methods

2.1 Microfluidic chip design and fabrication

The SFF microfluidic device creates the swirl effect by forcing the liquid through a circular blade array, con-centric to the exit channel and tangent to the liquid flow. Although the microchip design is 3D, the fabrica-tion process is not. Basically, it consists in a regular 2D engraving into one of the microfluidic chip slabs. Two concentric 50 µm filters were placed prior to the blade array to homogenize and filter the liquid flow and to reinforce the rigidity of the chip structure. Based on the

Fig. 2 (a) Sketch of the swirl flow-focusing microfluidic de-vice. (b) Closer view displaying the inner filter and the cir-cular blade array. (c) Sketch of the imposed swirl leading to the formation of a stable gas meniscus that breaks into mo-nodisperse microbubbles.

numerical results obtained using 3D CFD simulations (FLUENT) the following blade parameters were selec-ted: the closest radial position of the blades to the exit channel R1 = 150 µm, the blade length L = 200 µm, its width W = 20 µm, the number of blades n = 8 and the pith angle α = 0◦, 40◦, 60◦ and 80◦. Rectangu-lar blades were chosen for fabrication simplicity. The blades height was equal to the height of the chamber H = 30 µm.

The device is made up of a glass wafer (100 mm diameter, 1.2 mm thickness, Borofloat 33; Schott Ger-many) containing the outlet hole and a silicon wafer (p-type, 5-10 Ohm cm resistivity, 100 mm diameter, 525 µm thickness, {100} crystal orientation; Okmetic Finland) which has both, the gas and the liquid inlet holes, and the microfluidic chamber as shown in Figure 2.

The L = 1.2 mm thick Borofloat glass wafer was pro-cessed by FEMTOprint SA (Switzerland) using their 3D microstructuring technique to create a hole with a diameter D = 80 µm, throughout the entire thickness of the glass wafer assuring a perfect alignment between the gas inlet and the emulsion outlet. The use of glass has a double purpose: (i) since L/D=15, it serves as a microbubble collection channel and (ii) it allows the transversal view of the exit channel.

On the silicon wafer, a 500 nm thick silicon oxide layer was grown by wet oxidation. Subsequently, the pattern of the microfluidic chamber was transferred via standard photolithography and plasma etching into the silicon oxide layer (Adixen AMS100; Adixen France). The photoresist was removed and a new photolitho-graphy step was performed with the mask containing the pattern of the inlet holes. Using deep reactive ion et-ching (Bosch process, Adixen AMS100; Adixen France) the 80 µm inlet holes were etched into the silicon un-til the silicon oxide on the backside of the wafer was reached. The photoresist was removed and the remai-ning silicon oxide layer was used as a hard mask to deep reactive ion etch the H = 30 µm deep microfluidic chamber. After cleaning, the silicon oxide was remo-ved by etching in 50% hydrofluoric acid and the wa-fer was oxidised a second time with the newly formed 1 µm thick oxide layer striped afterwards. These last steps were performed in order to remove any residual silicon structures smaller than 1 µm, which might have remained at the location of the inlet holes due to the combination of the two deep reactive ion etching steps. Prior to bonding, the glass wafer and the silicon wa-fer were cleaned in a Piranha solution for 15 minutes. The wafer pair was aligned in a mask aligner (EV620 maskaligner; EVG Austria) and the anodic bonding was performed in a vacuum at 400◦ C for 1 hour with 800 Volt applied (EV-501 Anodic Bonder; EVG Austria). As a final step, the bonded wafer stack was diced into chips (Disco DAD 321, Disco Japan) with adhesive foil protecting the in and outlets from contamination.

2.2 Experimental setup

The swirl flow-focusing microfluidic device was moun-ted on a xyz stage for precise translation. A high-speed camera (Shimadzu HPV2) with a resolution of 312×260 px2 _{when operated at an acquisition rate of 1 Mfps,}

combined with a flash light source (WalimexPro VC600), was placed perpendicularly to the glass outlet channel.

The continuous phase was Milli-Q water. The sur-face tension between air and water σ was lowered from 72 to 40 mN/m by adding a 2% (w/v) of Tween 80

(Sigma Aldrich) to the water. The liquid flow rate Qlis

controlled by means of a high-precision syringe pump (Model 11 Plus, Harvard Apparatus). The precise con-trol of the gas flow rate Qgrequires imposing a pressure

gradient pg through a pressure regulator (11-818-100,

Norgren) and is measured using a digital manometer (Digitron 2030P). To avoid fluctuations in the gas flow rate the air was injected through a fused silica tube with 0.2 m in length and 75 µm of inner diameter. In order to prevent channel clogging by dust particles, an in-line filter (Parker, 0.01 µm ) was added to the gas flow line. The swirl flow-focusing chip is connected to the gas and liquid flow lines using peek Nanoports (As-semblies, Nanoport, Upchurch Scientific).

The bubble diameter dband bubbling frequency fb

of at least 100 images, are measured via image proces-sing (ImageJ). Knowing db and fb, the volumetric gas

flow rate is determined as Qg = πd3bfb/6. For low gas

to liquid flow rate ratios, where bubbles tend to have a more elongated shape, the equivalent diameter was likewise calculated.

3 Results and discussion

Monodisperse bubbles (polydispersity index PDI∼ 5%) ranging in size between 6 µm and 110 µm and at a high production rate (fb ∼ 105 Hz) were generated.

Bub-bles with diameters below 13 µm experience a higher PDI due to the high-speed camera resolution. We ac-complished ∼ 1000 experimental points varying the li-quid flow rate from Ql = 0.5 ml/min to Ql = 1.75

ml/min and selecting gas pressures from pg= 200 mbar

to pg = 2300 mbar, corresponding to flow rate ratios

between Qg/Ql= 0.01 and Qg/Ql = 1.

The chosen liquid flow rate range covered both, the bubbling regime (lower values) and the jetting regime (higher values). In presence of high liquid flow rates, specially above Ql = 1.75 ml/min, bubble jet

forma-tion becomes increasingly susceptible to perturbaforma-tions, preventing the meniscus formation due to gas comsibility effects and hydrodynamic feedback. Gas pres-sure was selected to enpres-sure bubbling and was gradually decreased until no bubbles were ejected.

For the range of liquid flow rates investigated here, Re = ρlvbD/µl ∼ O(700), being ρl and µl the liquid

density and viscosity, respectively, and vb the velocity

of the bubbles at the exit channel. This estimation in-dicates that the flow at the exit channel is laminar. Figure 3 shows the effect of increasing the liquid flow rate Ql for a 60◦ blade SFF microfluidic device and

a constant value of the gas pressure pg confirming the

VoF numerical predictions of Herrada and Ga˜n´an-Calvo (2009). Increasing the water flow rate results in smaller

a)

b)

c)

85 mm Fig. 3 Series of images showing the effect of increasing the liquid flow rate for a 60◦blade swirl flow-focusing microfluidic device and a constant value of the gas pressure: (a) Ql= 0.5

ml/min, pg = 851 mbar, db = 77.03 µm, fb = 1.01× 104 Hz; (b) Ql = 1 ml/min, pg = 876 mbar, db = 36.57 µm, fb = 8.11× 104 Hz; (c) Ql = 1.5 ml/min, pg = 847 mbar, db= 13.17 µm, fb= 2.92× 105Hz. a) b) c) 85 mm Fig. 4 Series of images showing the effect of increasing the swirl flow-focusing microfluidic device blade angle for a cons-tant gas to liquid flow rate ratio Qg/Ql = 0.07: (a) 0◦,

Ql = 1 ml/min, db = 45.15 µm, fb = 2.42× 104 Hz; (b)

40◦, Ql= 1.25 ml/min, db= 28.64 µm, fb= 1.19× 105 Hz;

(c) 60◦, Ql= 1.5 ml/min, db= 23.72 µm, fb= 2.5× 105Hz.

bubbles and higher breakup frequencies but also na-rrows the Qlrange where monodisperse bubbles can be

generated. The same trends were experimentally obser-ved for all the SSF microchip blade angles. Bubbles of 6 µm in diameter at a production rate of 2.14× 105_Hz

can be obtained when the 60◦ blade SFF microdevice is used under the appropriated operating conditions.

Figure 4 displays the effect of increasing the SFF microchip blade angle for a constant value of the gas to liquid flow rate ratio Qg/Ql. Accordingly to the VoF

nu-merical simulations presented by Herrada and Ga˜n´ an-Calvo (2009), larger values of the SFF microchip blade angle results in smaller bubbles and higher breakup fre-quencies but also restricts the Qlrange where

monodis-perse bubbles can be generated.

In order to have a reference case for the comparison between different blade angles, a 0◦SFF device was fa-bricated. If a conventional FF device is compared with

6003

n

4

003µH0 ×

4

0040 ×

4

004µ8H0 ×

4

004µH0 ×

4

004µ6H0 ×

G

0i





n

6003µdc

n

6004µ6b

n

G

0i





n

I0in0

600Hµz6

n

I0in0

Fig. 5 Bubble diameter versus microbubble production fre-quency for different blade angles. The indicated swirl factor

S is the tangent of the blade angle in this work. The color

codes for each liquid flow rate are also indicated.

a SFF, the breakup mechanism and the pressure drop might be completely different and the achieved reduc-tion in bubble size could not be correctly accounted for. A reduction up to 57% in the microbubble diameter is accomplished when compared to a 0◦ SFF.

Figure 5 shows the bubble diameter db(µm) versus

microbubble production frequency fb(Hz), for different

blade angles. The effect of the swirl is clearly shown in this figure: the swirl factor is defined as S = tan(α). This factor is expected to be approximately concurrent with that defined in Herrada and Ga˜n´an-Calvo (2009), since the blade angle α approximately defines the ra-tio of tangential to radial speeds. According to conser-vation of mass and angular momentum (excluding the many different boundary layer effects taking place), this ratio is expected to be approximately conserved when the liquid is eventually discharged through the outlet channel. In the absence of swirl (Figure 5(a)) the bubble size is modestly reduced by increasing the frequency of bubbling (i.e. reducing the gas flow rate). In contrast, a drastic decrease in the bubble size can be observed in the presence of swirl (Figure 5(b)) around a critical

frequency range, indicating a transition associated to the stabilization of the gas meniscus and the presence of jetting. This is coincident with what was described in Herrada and Ga˜n´an-Calvo (2009). One can also ob-serve that the increase of swirl (figures 5(c) & (d)), or the liquid flow rate over certain levels do not necessarily afford much better results in terms of a clear and repro-ducible decrease of bubble size, owing to the increase flow instabilities and incipient turbulence.

To further represent our results in the framework of prior physical understanding, Figure 6 depicts the microbubble diameter normalized with the exit chan-nel diameter db/D as a function of the gas to liquid

flow rate ratio Qg/Qlfor different SFF microchip blade

angles (0◦, 40◦, 60◦ and 80◦) and three representative liquid flow rates (Ql=0.5, 1 and 1.5 ml/min).

The figure shows two perfectly differentiable trends related to the two existing regimes. The experimental points corresponding to a bubbling regime, where no stable meniscus is created, are in the upper blue re-gion, above the black solid line. By contrast, the expe-rimental points where the combination of the swirl and flow-focusing effects enables the formation of a stable meniscus are in the lower green region, following the black solid line. In the latter situation, bubbles consi-derably smaller than the characteristic geometric length are ejected.

The black solid line in Figure 6 represents the sca-ling law presented by Ga˜n´an-Calvo (2004) using con-ventional axisymmetric flow-focusing devices

db/D = η (Qg/Ql)0.4, (1)

where η = 1.1 is a universal constant. In this work, we have found a slight deviation (η = 0.9) in the coefficient proposed by Ga˜n´an-Calvo (2004) due to a vena con-tracta effect. The high aspect ratio between the chip chamber height H and the exit channel diameter D, combined with the sharp edge of the entrance of the exit channel leads to a smaller effective exit diameter. The inclusion of the two colored regions constitutes a visual help to easily distinguish between the bubbling and the jetting regimes. The intent is not to give an exact sepa-ration (since transitions are never neat) but to be an eye guide for the reader. The straight boundary is chosen accordingly to the power-law fit proposed by Ga˜n´ an-Calvo (2004) which is followed by our experiments. The same boundary is used for the three plots in Figure 6. The figure also manifests that both regimes, bubbling and jetting, can be obtained for a particular gas to li-quid flow rate ratio and a given SFF microchip blade angle. Not only is gas to liquid flow rate ratio (Qg/Ql)

Fig. 6 Dimensionless bubble diameter versus gas to liquid flow rate ratio for different microchip blade angles and increasing values of the liquid flow rate: a) Ql= 0.5 ml/min; b) Ql= 1 ml/min; c) Ql= 1.5 ml/min. The black solid line is the power-law

fit proposed by Ga˜n´an-Calvo (2004). The upper blue region corresponds to a bubbling regime whereas the lower green region relates to a jetting regime. The colored regions are not an exact boundary but an eye guide for the reader to easily distinguish between the bubbling and the jetting regimes.

relevant to reach a jetting regime, but also their abso-lute flow rates (Qg and Ql). For instance, in Figure 6

(c) for a gas to liquid flow rate ratio Qg/Ql = 0.1 we

have two completely different experimental points: (i) a microbubble produced in the bubbling regime with a dimensionless bubble diameter db/D = 0.57, a gas

flow rate Qg= 0.15 ml/min, a liquid flow rate Ql= 1.5

ml/min and a production rate fb= 4.16× 104 Hz and

(ii) a microbubble produced in the jetting regime with a dimensionless bubble diameter db/D = 0.37, a gas

flow rate Qg= 0.14 ml/min, a liquid flow rate Ql= 1.5

ml/min and a production rate fb= 1.42× 105Hz. The

transition between both regimes is a delicate boundary that can be crossed with a change in gas pressure of just a few milibars. This sensitivity to small changes is especially present for a blade angle equal to 0◦. As soon as the blade angle increases, the possibility of two different regimes for similar operating conditions disap-pears, which strongly reinforces the convenience of the swirl flow-focusing configuration.

The transition from bubbling to jetting regime can be accomplished (for a fixed geometry and fluids) by increasing the liquid flow rate, as it was previously seen in Figure 3. Our experimental study demonstrates that this transition can also be conducted imposing a swirl on the liquid. Both effects are comprised in Figure 6: the increase of liquid flow rate and microchip blade angle. Furthermore, the stronger the swirl is, i.e. higher the blades angle, the lower liquid flow rates are needed to work on jetting conditions. Thus, the shift to jetting is reached more easily thanks to the imposed swirl. The transition occurs around Ql = 1.5 ml/min using 0◦

a)

b)

SFF

85 mm Fig. 7 (a) Numerical simulation illustrating an undulating bubble train. (b) Image showing the experimental undulating bubble train for a 40◦blade angle SFF microdevice, Ql= 1.75

ml/min, pg= 1341 mbar, db= 18.45 µm, fb= 3.6× 105Hz.

blade angle chips, but decreases to 1.25 ml/min for 40◦ blade angle microchips and to 1 ml/min for 60◦ and 80◦ microdevices. We observed a plateau on the swirl general performance above 60◦ blade angle, related to the previously observed drastic decrease of the effect of swirl over an optimal strength (see Herrada and Ga˜n´ an-Calvo (2009)). Part of that decrease could be attributed to the enhancement of perturbations.

Figure 7 shows the bubble train undulation predic-ted by the VoF numerical simulations and its experi-mental observation. The simulation was previously ob-tained by Herrada and Ga˜n´an-Calvo (2009) for pure water (without surfactant) and imposing numerically a swirl on a conventional axisymmetric Flow-Focusing

device. This undulation is due to experimental pertur-bations, such as dust particles, that may disrupt the axial symmetry of the jet. It also confirms that there is no bubble size change despite the trajectory deviation, which was forced in the simulations and naturally oc-curred in some experimental cases. The helicoidal trace followed by the bubbles is the result of the imposed li-quid swirl, migrating the bubbles towards the regions of lower pressure at the axis of the liquid vortex.

Lastly, experiments with ultra-pure MiliQ water (wit-hout surfactant) were also conducted. The results sho-wed an overall similar behavior, although the increase of surface tension favored the presence of hydrodynamic feedback. Consequently, the accessible flow rate ratios range was drastically reduced.

4 Conclusions

A novel swirl flow-focusing microfluidic device for the production of monodisperse microbubbles at a high pro-duction rate is presented. The swirl effect is achieved thanks to the rotation induced by a circular array of mi-croblades turned a certain angle. The addition of a swirl component into the focusing liquid stabilizes the gas meniscus from which a steady gas ligament issues brea-king into monodisperse microbubbles. Furthermore, the swirl is shown to expand the bounds of the jetting mode inhibiting the dripping mode. As consequence of the ex-tension of the jetting regime, a reduction up to 57% in the microbubble diameter is accomplished when com-pared to conventional axisymmetric flow-focusing mi-crodevices.

Inspired by the numerical results of Herrada and Ga˜n´an-Calvo (2009) and Herrada et al (2011), we per-form 3D CFD simulations (FLUENT) to determine the most promising geometric dimensions for the swirl flow-focusing microdevices. Based on the simulations, sili-con microchips with blade angles 0◦, 40◦, 60◦ and 80◦ were fabricated. Working under the appropriate experi-mental conditions, monodisperse bubbles (PDI ∼ 5%) ranging in size between 6 µm and 110 µm and at high production rate (fb ∼ 105 Hz) can be generated. An

exhaustive experimental study (∼ 1000 experimental points) is performed validating previous VoF numeri-cal simulations and complying with the flow-focusing scaling law proposed by Ga˜n´an-Calvo (2004). The 60◦ swirl flow-focusing microfluidic device shows the best performance, among our tested chips, with a trade off between swirl effect and robustness against perturba-tions.

Acknowledgements The authors would like to acknowledge financial support from Spanish Government Ministry MEIC

and Regional Government under the Contract DPI2013-46485 and P11-TEP-7465, respectively. They would also like to ack-nowledge the technical assistance of S. Schlautmann in the fabrication of the microfluidic devices and Manuel Gonz´alez and Jorge L´opez for their technical assistance during the se-tup preparation.

References

Anna SL, Bontoux N, Stone HA (2003) Formation of dispersions using flow focusing in microchannels. Ap-plied Physics Letters 82(3):364–366

Castro-Hern´andez E, van Hoeve W, Lohse D, Gordi-llo JM (2011) Microbubble generation in a co-flow device operated in a new regime. Lab on a Chip 11(12):2023–2029

Dollet B, Van Hoeve W, Raven JP, Marmottant P, Versluis M (2008) Role of the channel geometry on the bubble pinch-off in flow-focusing devices. Physi-cal Review Letters 100(3):034504

Ferrara K, Pollard R, Borden M (2007) Ultrasound mi-crobubble contrast agents: Fundamentals and appli-cation to gene and drug delivery. Annual Review of Biomedical Engineering 9:415–447

Ga˜n´an-Calvo AM (2004) Perfectly monodisperse mi-crobubbling by capillary flow focusing: An alternate physical description and universal scaling. Physical Review E 69(2):027301

Ga˜n´an-Calvo AM, Gordillo JM (2001) Perfectly mo-nodisperse microbubbling by capillary flow focusing. Physical Review Letters 87(27):274501

Ga˜n´an-Calvo AM, Herrada MA, Garstecki P (2006) Bubbling in unbounded coflowing liquids. Physical Review Letters 96(12):124504

Garstecki P, Gitlin I, Diluzio W, Kumacheva E, Stone HA, Whitesides GM (2004) Formation of monodis-perse bubbles in a microfluidic flow-focusing device. Applied Physics Letters 85:2649–2651

Garstecki P, Ga˜n´an-Calvo A, Whitesides G (2005) For-mation of bubbles and droplets in microfluidic sys-tems. Technical Sciences 53(4)

Garstecki P, Fuerstman MJ, Stone HA, Whitesides GM (2006) Formation of droplets and bubbles in a micro-fluidic t-junction˜nscaling and mechanism of break-up. Lab on a Chip 6(3):437–446

Herrada MA, Ga˜n´an-Calvo AM (2009) Swirl flow fo-cusing: A novel procedure for the massive produc-tion of monodisperse microbubbles. Physics of Fluids 21(4):042003

Herrada MA, Ga˜n´an-Calvo AM, L´opez-Herrera JM (2011) Generation of small mono-disperse bubbles in axisymmetric t-junction: the role of swirl. Physics of Fluids 23(7):072004

Kennedy J, Wu F, Ter Haar G, Gleeson F, Phillips R, Middleton M, Cranston D (2004) High-intensity fo-cused ultrasound for the treatment of liver tumours. Ultrasonics 42(1-9):931–935

Marmottant P, Van Der Meer S, Emmer M, Versluis M, De Jong N, Hilgenfeldt S, Lohse D (2005) A mo-del for large amplitude oscillations of coated bubbles accounting for buckling and rupture. Journal of the Acoustical Society of America 118(6):3499–3505 Takahashi M (2005) Potential of microbubbles in

aqueous solutions: Electrical properties of the gas - water interface. Journal of Physical Chemistry B 109(46):21858–21864

Wei K, Jayaweera A, Firoozan S, Linka A, Skyba D, Kaul S (1998) Quantification of myocardial blood flow with ultrasound-induced destruction of micro-bubbles administered as a constant venous infusion. Circulation 97(5):473–483

Whitesides G (2006) The origins and the future of mi-crofluidics. Nature 442(7101):368–373